+1 862 207 3288 WEEK 5 ASSIGNMENT:
HAND-IN ASSIGNMENT
• Briefly explain the concept of incremental cost-effectiveness ratio (ICER).
• Using the data below, calculate the average and incremental ICERs. Assume that the options are mutually exclusive. Hint: Sort options according to cost.
• Comment about the dominance of Beclomethasone over Salmeterol-high dose.
• Indicate your choice when the 3-month budget is USD 200 and USD 300.
• If Ipratropium treatment is the existing one, and if your incremental threshold ratio is USD 500 per FEV1 unit, what are the feasible options?
Cost (USD) over 3-month study period Change in FEV1 (l)
Placebo 60.00 1.1
Salmeterol: low dose 170.00 1.5
Salmeterol: high dose 360.00 1.6
Ipratropium 90.00 1.3
Beclomethasone 280.00 1.7
WEEK 5 INTRODUCTION
WEEKLY SUMMARY
Last week you were introduced to health financing. You explored mechanisms, objectives, and activities of health financing. Implicit and explicit priority-setting approaches were discussed. The aim of this week is to introduce you to economic evaluations. The main types of economic evaluation will be explained in this session. You will calculate and interpret incremental cost-effectiveness ratios as well as apply simple forms of discounting. The main approaches of sensitivity analysis will be explained.
OBJECTIVES
Learning objectives for Week 5
On completion of this week, students will be able to:
• Describe the key types of economic evaluation.
• Calculate and interpret incremental cost-effectiveness ratios.
• Identify the main types of costs described in economic evaluation.
• Explain what is meant by discounting in economic evaluation.
• Understand the role of sensitivity analysis in economic evaluation.
WEEK 5 LEARNING RESOURCES
REQUIRED RESOURCES
• Morris, S., Devlin, N. & Parkin, D. (2012) ‘Chapter 10: Principles of economic evaluation in health care’. In: Economic analysis in health care. 2nd Ed. Chichester, England: John Wiley & Sons, pp. 232-252.
• Morris, S., Devlin, N. & Parkin, D. (2012) ‘Chapter 12: Economic evaluation methods’. In: Economic analysis in health care. 2nd Ed. Chichester, England: John Wiley & Sons, pp. 286-320.
• Palmer, S., Byford, S. & Raftery, J. (1999)‘Types of economic evaluation’, British Medical Journal, 318 (7194), p. 1349 [Online]. Available from: http://search.proquest.com.ezproxy.liv.ac.uk/docview/204012213/4B8DDC0EC0AB4ABCPQ/1?accountid=12117 (Accessed: 13 June 2014).
• Raftery, J. (1998) ‘Economic evaluation: an introduction’, British Medical Journal, 316 (7136), pp. 1013-1014 [Online]. Available from:
• http://search.proquest.com.ezproxy.liv.ac.uk/docview/203991216/B38E1D7447504736PQ/2?accountid=12117 (Accessed: 18 August 2010).
• Byford, S. & Raftery, J. (1998) ‘Perspectives in economic evaluation’, British Medical Journal, 316 (7143), pp. 1529-1530 [Online]. Available from:
• http://search.proquest.com.ezproxy.liv.ac.uk/docview/203932400/B2B1838AE564314PQ/1?accountid=12117 (Accessed: 13 June 2014).
SUPPLEMENTAL RESOURCES
• Evans, D.B., Edejer, T.T.-T., Adam, T. & Lim, S.S. (2005) ‘Methods to assess the costs and health effects of interventions for improving health in developing countries’, British Medical Journal, 331 (7525), pp. 1137-1140 [Online]. Available from:
• http://www.bmj.com.ezproxy.liv.ac.uk/content/331/7525/1137.full (Accessed: 18 August 2010).
• Ginnelly, L., Sculpher, M., Bojke, C., Roberts, I., Wode, A. & Diguiseppi, C. (2005). ‘Determining the cost effectiveness of a smoke alarm give-away program using data from a randomized controlled trial’, The European Journal of Public Health, 15 (5), pp. 448-453 [Online]. Available from:
• http://eurpub.oxfordjournals.org/content/15/5/448.long (Accessed: 18 March 2014).
• Phillips, C. & Thompson, G. (2009) ‘What is cost-effectiveness?’, What Is…? Hayward Medical Communications [Online]. Available from:
• http://www.medicine.ox.ac.uk/bandolier/painres/download/whatis/Cost-effect.pdf (Accessed: 18 August 2010).
• Center for Evidence-Based Medicine(n.d.) Glossary of EBM terms [Online]. Available from: http://ktclearinghouse.ca/cebm/glossary/ (Accessed:17 June 2014).
WEEKLY NOTES
Week 5: Costing and cost-effectiveness analysis
INTRODUCTION
Last week you were introduced to health financing. You were able to learn about mechanisms, objectives, and activities of health financing. Implicit and explicit priority setting approaches were discussed.
The aim of Week 5 is to introduce you to economic evaluations. The main types of economic evaluation will be explained in this session. You will learn how to calculate and interpret incremental cost-effectiveness ratios (ICERs) as well as how to apply simple forms of discounting. The main approaches of sensitivity analysis will be explained.
The core readings for this week are:
• Morris, S., Devlin, N. & Parkin, D. (2012) ‘Chapter 10: Principles of economic evaluation in health care’. In: Economic analysis in health care. 2nd ed. Chichester, England: John Wiley & Sons, pp. 232-252.
• Morris, S., Devlin, N. & Parkin, D. (2012) ‘Chapter 12: Economic evaluation methods’. In: Economic analysis in health care. 2nd ed. Chichester, England: John Wiley & Sons, pp. 286-320.
INTRODUCTION TO ECONOMIC EVALUATION
What is economic evaluation?
There are four main methods of economic evaluation in healthcare. Each of the evaluation methods assesses the costs and benefits of alternative healthcare interventions. The main difference between the four methods lies in the way that they deal with health benefits. These methods of economic evaluation of healthcare interventions are: cost-minimisation analysis, cost-effectiveness analysis, cost-utility analysis, and cost-benefit analysis.
Origins of economic evaluation
The four methods of economic evaluation that are commonly used today for the evaluation of healthcare interventions in health economics have their origins in cost-benefit analysis (CBA). CBA was developed over 50 years ago within the framework of welfare economics to help with public-sector investment planning. In the private sector, costs, prices, and profits can be used to guide decisions about private investment. Goods and services provided by the public sector are often provided free at the point of consumption or at substantially below their production costs. CBA was developed as a public-sector alternative to private-sector profit-and-loss accounting. During the 1930s, the first CBAs were carried out in the US (flood control and transport investment programmes). It was not until the 1960s that CBA was introduced in the UK to help plan public-sector transport investment such as in the building of the M1 motorway.
Why is economic evaluation important in healthcare?
The use of economic evaluation is becoming increasingly frequent in the assessment of healthcare technologies throughout the world. Where trade-offs are likely between costs and outcomes—or between different types of outcomes—the use of economic evaluation techniques can offer a systematic, objective, and value-free means of evaluation. Increasingly, healthcare decision-makers are relying on economic evaluations in their day-to-day work. Adopting a comprehensive approach to the identification, measurement, and valuation of costs and benefits aims to ensure value for money from scarce healthcare resources.
Possible outcomes of economic evaluation studies
Typically, economic evaluation requires the identification, measurement, and valuation of the costs and benefits of a ’new’ technology compared to an ’old’ technology. There are four possible outcomes of this comparison.
• Both the costs and the benefits are higher for the new healthcare technology; this means that incremental cost-effectiveness analysis (CEA) needs to be carried out.
• Costs are lower and health benefits are higher for the new healthcare technology; this means that the new technology dominates and is the preferred option.
• Costs are lower with reduced health benefits for the new healthcare technology; this means that incremental CEA needs to be carried out.
• Costs are higher and health benefits are lower for the new healthcare technology; then this technology is inferior and is not the preferred option.
The following articles present basic details of economic evaluations:
• Palmer, S., Byford, S. & Raftery, J. (1999)‘Types of economic evaluation’, British Medical Journal, 318 (7194), p. 1349 [Online]. Available from: http://search.ebscohost.com.ezproxy.liv.ac.uk/login.aspx?direct=true&db=mnh&AN=10323827&site=ehost-live&scope=site (Accessed: 18 August 2010).
• Raftery, J. (1998) ‘Economic evaluation: an introduction’, British Medical Journal, 316 (7136), pp. 1013-1014 [Online]. Available from:
• http://search.ebscohost.com.ezproxy.liv.ac.uk/login.aspx?direct=true&db=mnh&AN=9550969&site=ehost-live&scope=site (Accessed: 18 August 2010).
• Byford, S. & Raftery, J. (1998) ‘Perspectives in economic evaluation’, British Medical Journal, 316 (7143), pp. 1529-1530 [Online]. Available from:
• http://search.ebscohost.com.ezproxy.liv.ac.uk/login.aspx?direct=true&db=mnh&AN=9582152&site=ehost-live&scope=site (Accessed: 18 August 2010).
COSTS IN ECONOMIC EVALUATION
In health economics, costs can be defined and classified in a variety of ways. Unfortunately, economists have sometimes used words in slightly differently ways from accountants, which can lead to confusion. The following discussion of costs is relevant to the identification, measurement, and valuation of costs in any method of economic evaluation.
Opportunity cost denotes a concept in which cost is more than simply the money we spend to obtain something. It is the opportunities and benefits that we give up to obtain a thing that is the real cost. If we spend money, we must forgo all the other things that we might have done with that money and the benefits we could have consumed. For example, if we convert an empty ward into a recreation area, we need to consider what other uses that ward might have had. Opportunity costs contrast with the accounting definition of cost, as accountants are only interested in the amount of money spent during a specific period of time.
It is important to distinguish whether the costs under consideration relate to the total amount of care provided or whether they relate to changes in the amount of care provided (the distinction between average costs and marginal costs). Marginal cost is the change in extra cost of producing one more unit of output—the extra cost of increasing activity by one unit. Average cost is the total cost of a good or service divided by the number of units of output produced. Most healthcare policy decisions are not concerned with whether a certain service should—or should not—be provided; rather, they are concerned with whether the service should be expanded, contracted, or remain at prevailing levels of provision. For example, a purchaser may need to decide whether existing levels of provision of coronary artery bypass grafts are appropriate, should be expanded, or should be contracted.
All such decisions should be based on marginal costs—i.e., the difference in total cost of treating one more patient—rather than on the average cost of providing a service. A somewhat different application of the concept of marginal cost arises, for example, when considering how long a patient remains in hospital. On the first day the patient might be doing quite poorly, particularly if they have an operation on that day. There would then be a lot of costs associated with the operation and the post-operative care. A few days later the patient might be mobilised and looking after themselves to a much greater extent. Towards the end of a patient’s hospital stay, the cost of an additional day (the marginal cost of an extra day in hospital) may be quite small.
Incremental costs reveal the cost of switching from one treatment option to an alternative treatment option and are sometimes confused with marginal costs in the published literature.
Organisational costs
• Classification of costs according to traceability of costs
Direct costs are directly traceable to a cost object. For a chest x-ray film (the cost object) the cost of material, consumables, and that of the technologist (cost of time spent on the process) are some of the direct costs. Indirect costs/overheads cannot be traced directly to a cost object. These costs are sometimes called overheads. For a chest x-ray film, the cost of electricity and water are some of the indirect costs.
• Classification of costs according to their behaviour
Fixed costs do not vary with changes in the level of activity. Building and equipment costs and salaries of staff are examples for fixed costs. Variable costs vary with changes in the level of activity. Costs of drugs and consumables are examples of variable costs.
• Classification of costs according to financial accounting
Capital costs relate to the purchase of items that can be used for several years. These costs are usually incurred at one time point, whereas the benefits accrue over the useful life of the good. Because capital investments for such items as equipment can be used over the duration of a programme, the costs of capital expenditures may be spread out over that time period. If we are considering an acute condition that requires the use of a piece of capital equipment, the cost of that equipment will have to reflect the fact that it is expensive to ‘tie up’ money in the capital equipment for a long time (or alternatively borrow money and repay the interest to finance it).
Recurrent costs relate to short-term expenses such as salaries, travel costs, purchase of items that can only be used for less than a year, etc.
• Costing methods
In absorption costing, the products/cost objects are charged with all the direct costs and a share of the overheads is absorbed. In marginal costing the cost is estimated by dividing the costs into two categories: fixed or variable.
Activity-based costing is a technique of establishing costs more accurately. Activities that are needed to produce a certain product or service are viewed as giving rise to costs. As an example, the cost of the activity of registering patients can be assigned to each patient based on the time taken by staff for the registration.
Costs other than organisational costs
Direct non-medical costs refer to the costs of patients and their families as they consume healthcare (e.g., over–the-counter medicines or cost of transport to and from hospital). Indirect non-medical costs refer to the value of lost productivity or lost earnings from time off work due to illness. Intangible costs refer to the social and psychic costs of treatments (e.g., pain and stress). These costs are very difficult to describe, and techniques are being refined to measure these costs. Even if intangible costs are not quantified, it is good practice to explicitly record the fact that they exist.
Factors which also influence costs in economic evaluation
(i) Perspective (or viewpoint)
A key point to consider in any economic evaluation is the perspective from which the study is conducted, i.e., the group(s) whose costs we wish to include in the analysis. In general, the societal perspective is considered the most appropriate as it includes analysis of all costs and benefits that are incurred by the programme or intervention. However, a decision-maker faced with a limited budget has a strong incentive to ignore the societal view and concentrate entirely on the costs that fall within his own budget. The perspective chosen may be a crucial determinant of the decision made. For example, a study of therapeutic options for the relief of back pain that restricts the analysis to costs imposed on the health service might conclude that treatment is costly and of limited benefit. In contrast, the same study adopted from a societal perspective would also incorporate the wider benefits arising to society such as a reduction in working days lost and the patients’ enhanced ability to undertake activities of daily living. Given this wider perspective, it is highly likely that the cost-benefit ratio identified would be more favourable. Do you think it is practical to estimate costs from a societal perspective?
The National Institute for Health & Clinical Excellence (NICE), which recommends treatments for England and Wales, uses the perspective of costs to the health service and personal social services when evaluating the cost-effectiveness of treatments. Do you think this is fair, or do you think it would be fairer to use the costs to society?
(ii) Choice of comparator
In any economic evaluation, choosing an appropriate comparator(s) requires a great deal of consideration, as again it may fundamentally influence the results of the analysis. The comparator should be the current gold-standard treatment. If not, then an explanation of why this is not the case is required. Where pharmacological interventions are being assessed, scrutiny of the appropriateness of the comparator drug and dosages is necessary. One drug may appear to be more effective than another simply because an old generation of drug or a sub-optimal dosage is being administered. A placebo is not usually an appropriate comparator, and it must be emphasised that a do-nothing alternative is not a no-cost alternative.
Return to top
DISCOUNTING IN ECONOMIC EVALUATION
There is often a difference in timing between the investment of health service resources and the associated health gain. In economic evaluation, discounting is used to reflect the fact that a cost or benefit occurring in the future is believed to be less valuable than the same cost or benefit occurring now. In general, we prefer to receive benefits now and pay costs in the future. Most people who have the choice of either paying a charge of £100 today or in 5 years’ time would prefer to pay the charge in 5 years’ time. Likewise, most people would prefer to receive a health improvement today rather than wait for an improvement in 5 years’ time. We use discounting to allow us to compare projects with different time profiles of costs and benefits (e.g., costs may occur in the first 3 years, whilst benefits accrue over a 10-year period).
Discounting allows us to express the future costs and benefits of projects in terms of their value today (present value). We do this by applying a discount rate to all future costs and benefits, i.e., any costs of benefits that are incurred beyond an initial 12-month period.
The following equation is used to discount costs and benefits:
P = ? Fn (1 + r) – n
where P = present value; Fn = future cost at year n; and r = discount rate.
In any economic evaluation, it is useful to present costs and benefits undiscounted in the first instance and then use sensitivity analysis to demonstrate how changing the discount rate can impact on the size of the cost-effectiveness ratio (CER). A higher discount reduces the weight of future costs and benefits.
There is health economics research devoted to exploring the use of discounting within the framework of economic evaluation. The major debate is around whether or not health benefits should be discounted. Proponents of discounting health benefits argue that if costs are discounted, then it is illogical not to discount benefits; i.e., if benefits are not discounted, then postponing a project becomes very attractive, as the costs will be ever-decreasing yet the level of benefits stays the same. This is another way of saying that only discounting costs has the effect of reducing the impact of such costs, potentially showing interventions to be more cost-effective than they would otherwise appear.
The main argument against the discounting of health benefits appears to centre around the debate that that there is no real evidence to suggest that individuals do value current health states more than future health states. It is relatively easy to accept that £100 spent now is worth more than £100 spent in 5 years’ time, but how does one compare a healthy year now to a healthy year in 5 years’ time? Also, it is argued that preventative programmes are penalised by the discounting of health benefits, as their benefits typically manifest over time.
WHAT IS COST-MINIMISATION ANALYSIS?
Cost-minimisation analysis (CMA) is an appropriate evaluation method when the case for a clinical intervention has been established and the preventative, diagnostic, or therapeutic procedures or programmes under consideration are expected to have exactly the same outcome. Indeed, a prerequisite for conducting a CMA is that the healthcare technologies being compared are clinically equivalent—this means that prospective economic evaluations alongside clinical trials are never designed as CMAs because clinical equivalence cannot be assumed before the trial results are known.
CMAs are typically based on the results of randomised controlled trials. In CMA, the least expensive option is preferred. This method of economic evaluation might be appropriate for comparing the costs of day surgery vs. inpatient treatment for the repair of hernias if the clinical outcomes and patient satisfaction are confidently known to be comparable. Another appropriate use of the results of a CMA would be a decision to prescribe a generic drug instead of a brand-name drug, which should, in theory, achieve the same effect at less cost.
However, demonstrating that the health outcomes of two different technologies are identical (equivalent) can be both time-consuming and complex. Frequently, cost analyses are erroneously labelled as CMAs (i.e., costs are estimated without any consideration of whether benefits are similar), which can lead to uninformed decision making.
Cost-minimisation studies are not to be confused with cost-of-illness studies. Cost-of-illness studies simply estimate the size of the cost burden to society of a specific disease.
WHAT IS COST-EFFECTIVENESS ANALYSIS?
Cost-effectiveness analysis (CEA) properly refers to an evaluation where the health benefit can be defined and measured in natural units (e.g., years of life saved, number of fully immunised children, etc.). The health benefits described in a CEA are uni-dimensional. Cost-effectiveness compares therapies whose outcomes can be measured in the same natural units and measures costs in money. For instance, if the desired outcome is relief of severe reflux oesophagitis, consideration might be directed towards estimating the costs per patient relieved of symptoms using a proton pump inhibitor compared to those using H2 blockers. This method might also be appropriate for comparisons of the cost-effectiveness of two or more drugs for the control of HIV infection where costs are measured in monetary terms and outcomes in terms of number of HIV infections avoided.
CEAs can be performed both within and across clinical specialties as long as a common natural unit of outcome is used (e.g., cost per life year gained). CEA is the most commonly applied form of economic analysis in the heath economics literature, and it is commonly used in drug therapy.
It cannot be concluded from a CEA that a healthcare intervention is economically worthwhile (benefits are greater than costs). A CEA asks the question ‘how should we do things?’ not ‘should we do things?’. Until recently, it has been tacitly assumed that when CEA is carried out, it has already been established that the treatment objective has been worth meeting. However, current interest in the estimation of a decision-maker’s willingness to pay for a unit of effectiveness has led to more explicit considerations of worthiness.
Finally, there are two main criticisms of using CEA. First, all of the health effects are assumed to be valued identically (e.g., a life year gained is valued the same for a 4-year-old as it is for an 80-year-old; e.g., 20 life years saved for 1 person is the same as 1 life year saved for 20 people). Secondly, it is argued that healthcare interventions are rarely so similar as to be measurable in terms of a single effect.
INCREMENTAL COST-EFFECTIVENESS RATIOS
In any CEA, we end up with two (or more) different cost-effectiveness ratios (CERs). The calculation of the incremental CER (ICER) reveals the cost per unit of benefit of moving from one healthcare option to an alternative option. ICERs are calculated using the following equation:
ICER = (Cost of drug A – Cost of drug B)
(Effectiveness of drug A – Effectiveness of drug B)
Whether or not an ICER is acceptable in terms of cost-effectiveness is very much dependent on the healthcare technology being evaluated and the aims and objectives of the research. The question that must be asked is whether or not the ICER is higher or lower than the decision-maker is prepared to pay per unit of health benefit; this value is known as the decision-maker’s threshold willingness to pay, ceiling ratio, or cost-effectiveness threshold. Often, information on its value can be found in the published health economics literature. If the monetary value of the ICER is similar to other ICERs for widely adopted healthcare technologies in the disease area, then the ICER may be assessed as being acceptable.
When more than two programmes are being compared, the following steps can be taken to conduct the CEA:
1 Rank options in terms of increasing costs
2 Eliminate options for which another option is cheaper and more effective (simple dominance)
3 Eliminate options where a combination of two options is cheaper and more effective (extended dominance)
4 Calculate the incremental cost and incremental benefit for each of the remaining options with respect to the previous option on the list
5 Calculate ICERs for the remaining options
Net benefit approach (alternative to ICERs)
Rather than concentrating on the calculation and comparison of ICERs, the goal of the net benefit approach is to represent the worthiness of the healthcare technology with costs and benefits being expressed in the same outputs.
For example, in CEA, the ICER is essentially equal to the change in costs divided by the change in benefits; where the costs are measured in monetary terms and the outcomes are measured in natural units of effectiveness. However, it is useful to express costs and effects in the same units. In order to represent costs and effects in the same units, the net benefit approach is used. The net benefit approach relies on the use of the ceiling ratio (the threshold willingness to pay for a unit of effect) to convert effects into the same units as costs (net monetary benefit, NMB), and convert costs into the same units as effects (net health benefit, NHB). The decision rules of the net benefit approach are outlined in Box 1.
Box 1: Decision rules for net benefits
Net benefit rule Decision rule
NMB = Value of the ceiling ratio x ? effects – ? costs If NMB > 0: the project is considered to be cost-effective
NHB = ?E – (?C/value of the ceiling ratio) If NHB > 0: the project is considered to be cost-effective
Clearly, the most important value in the net benefit approach is the size of the ceiling ratio, as it may vary depending on the decision-maker.
ROLE OF SENSITIVITY ANALYSIS IN ECONOMIC EVALUATION
We are rarely certain about all of the components of an economic evaluation. Common contributors to our uncertainty include: the effectiveness of the intervention, the quantities of resources used, the prices we should assign to resources, or the time that patients must take off work. In some settings there may be no data available, or the data are imprecise (e.g., hospital per diem costs rather than patient-specific costs per day in hospital). Methodological controversy (e.g., should benefits be discounted at the same rate as costs?) can also be explored using sensitivity analysis. One way of dealing with uncertainty is to recalculate the estimated costs and/or effects using alternative plausible values for uncertain parameters and consider whether different values would significantly alter the conclusions of the evaluation. Such a process is called a ‘sensitivity analysis’ since it addresses the question of how ‘sensitive’ the results are to the values used.
The first step in any sensitivity analysis is to identify the parameters which are uncertain (e.g., average length of stay for a patient with tuberculosis). The second step is to specify the range over which we think the parameter might vary. To illustrate, if we used an average length of stay of 80 days for patients with tuberculosis, and we actually believe that this number lies somewhere between 70 and 90 days, then we specify this range of values. We must also specify why we think this range is appropriate. For example, plausible ranges could be supported by evidence from reviews of the clinical literature or from expert clinical opinion.
The third step is to recalculate the CERs based on combinations of best guess (80 days), most conservative (70 days), and least conservative (90 days) values. Based on the reworked calculations, robustness of the study results may or may not be confirmed. If there is no change or very little change in the results of the economic evaluation, then you can be quietly confident that, despite uncertainty around your parameters, study results can be considered robust. However, as there is no guidance to what variation in results is acceptable, true robustness is very much open to interpretation.
There are different types of sensitivity analysis. One-way sensitivity analysis addresses individual sources of uncertainty. Estimates for each parameter are varied one at a time in order to investigate the impact on study results. For example, we might recalculate our cost-effectiveness estimate using a higher nursing wage rate (see worked example below). If nurses only comprise a small proportion of our total costs, even a large difference in wages may have little effect. Although one-way sensitivity analysis is the simplest and most common approach to sensitivity analysis, there exists the risk that interactions between parameters may not be captured. Multi-way sensitivity analysis does this with several sources of uncertainty together. More sophisticated forms of sensitivity analysis can use assumptions about probability distributions or information about such distributions from data collected in the study.
Net monetary benefit (NMB) and its sampling uncertainty can be presented diagrammatically as a function of the threshold willingness to pay (Figure 1). The middle line on the figure shows the mean NMB. The threshold willingness to pay value at the point where the line crosses the x-axis indicates the mean ICER. At this particular point, the NMB is zero. The NMB value at the point where the line crosses the y axis equals the cost difference between the two options. At this particular point, the threshold willingness to pay is 0.
Figure 1: Net monetary benefit and sampling uncertainty
Cost-effectiveness acceptability curve (CEAC) can be used to express the probability that an intervention is more cost-effective than the comparator given the sampling data. In CEAC, the probability of an option being cost-effective is graphed on the y-axis, and the threshold willingness to pay is graphed on x-axis. CEAC is estimated based on the joint uncertainty of cost and effect. The cost-effective plane can be used to estimate the probability of an option being cost effective for alternative levels of threshold willingness to pay. Figure 2 presents the cost-effectiveness plane for an asthma treatment scenario.
Figure 2: Scatter-points derived from bootstrap analysis of cost-effectiveness
First, scatter-points are estimated based on results of boot-strap analysis. Boot-strapping is a non-parametric method whereby the results of the analysis are randomly resampled many times to give a distribution of possible CERs. For a given threshold level (oblique lines on the cost-effective plan), the probability of the option being more cost-effective is the proportion of scatter-points below the threshold willingness to pay line. For a threshold willingness to pay of USD1,000 per additional FEV1, the scatter-points which show that the option is cost-effective, are either 1) dominant—less costly and more effective (south-east quadrant); 2) more costly and more effective, but having an ICER below the threshold value; or 3) less costly and less effective, but having an ICER below the threshold value. Based on the proportion of scatter-points below the threshold line, CEAC can be graphed (refer to Figure 3).
Figure 3: Cost-effectiveness acceptability curve
There are limitations to the use of sensitivity analysis. The main limitation of the approach is that the analyst has total discretion over which parameters are considered to be uncertain, so some sensitivity analyses can be criticised for not being data driven. Secondly, as there is no guidance as to what degree of variation of study results is acceptable, the interpretation of sensitivity analyses is often described as arbitrary.
